from torch import nn

model = nn.Sequential(
    nn.Embedding(vocab_size, embed_dim),
    nn.Linear(embed_dim, hidden_dim),
)
# Architecture defined; weights initialized randomly or loaded later